Research shows that maternal depression has adverse effects on mother-infant attachment and subsequent infant development (Cogill, Caplan, Alexandra, Robson & Kumar, 1986). The mechanisms through which this comes about are unclear. Murray & Cooper (1997) suggest an impaired pattern of mother-infant communication is responsible. Within this, Murray proposes that maternal speech may be a key factor. This study constitutes a preliminary exploration into the mechanisms through which maternal depression effects mother-infant interaction in South Africa. 147 predominantly Xhosa-speaking mother-infant dyads that took part in a broader epidemiological study on post-partum depression in Khayelitsha (Cooper, Tomlinson, Swartz, Woolgar, Murray & Molteno, 1999) made up the subjects. Maternal depression was assessed according to the Structured Clinical Interview for DSMIV (SCID). Maternal speech recorded from standard, five-minute, face-to -face mother-infant interactions was translated and analysed according to a coding system developed by Murray (Murray, Kempton, Woolgar & Hooper, 1993). The speech of depressed mothers to their two month old infants was compared to the speech of non-depressed mothers on dimensions of focus, affect and agency, and the role of infant gender was assessed. Results revealed no significant group differences for depression. Maternal speech to male infants was found to hold significantly less ascription of agency than to female infants. Findings suggest that maternal speech may be too narrow a marker of maternal depression in this context and that broader indices are needed. It is recommended that future research control for measures of social adversity, factor in cultural and language particularities, and consider contextual aspects of mother-infant interaction / attachment processes, in investigating the mechanisms through which post-partum depression leads to negative infant outcome in the developing world.